Perception-Based Models for Intelligent Image/Video Retrieval

odecrackΤεχνίτη Νοημοσύνη και Ρομποτική

29 Οκτ 2013 (πριν από 3 χρόνια και 9 μήνες)

63 εμφανίσεις

P
erception
-
B
ased Models for Intelligent Image/Video
Retrieval


Supervisor: Assoc Prof Wang Lipo, EEE

(
http://www3.ntu.edu.sg/home/elpwang
)

Co
-
Supervisor: Assoc Prof Lin Weisi, SCE

(
ht
tp://www3.ntu.edu.sg/home/wslin)


An important challenge in building a multi
-
media search engine is to account for the
differences between human perception of relevance and ma
chine perception of relevance.
The latter usually considers overall similarity between the query image/video and those to
be retrieved, whereas the former may be sometimes difficult to describe and involves
semantic understanding of the images/videos. Per
ception
-
based

models
aim at providing
info
rmation

on where/which is important objects/patterns in an image/video

from
human
perception

point of view
. Visual attention and sensitivity/visibility are possible cues.
These may help "priority" learning, target
oriented classification, active vision, etc.

In
this project, we intend to study innovative perception
-
based models and their
incorporations into machine learning algorithms for intelligent image/video retrieval.